NPS vs CSAT vs CES: Which Customer Experience Metric Should You Use?

Most teams asking "NPS vs CSAT vs CES" are really asking two questions at once: which number should we put on the dashboard, and which one will tell us what to fix. Those are not the same question, and confusing them is why so many CX programs track a score for years without ever moving it.

The short answer is this: use the metric that matches the touchpoint. Net Promoter Score (NPS) measures the overall relationship and likelihood to recommend. Customer Satisfaction (CSAT) measures how satisfied someone was with a specific interaction. Customer Effort Score (CES) measures how hard a customer had to work to get something done. Pick the one that fits what you are trying to learn at that moment. But the metric you choose matters far less than understanding what drives it. All three are lagging indicators. The leading indicators are the themes buried in the open-text feedback attached to every score. Thematic exists to surface those drivers, which is the half of the question the metric debate leaves out.

Below is a straight comparison of the three metrics, when each one is the right choice, and why the score you pick is only as useful as your ability to explain what is moving it.

The honest framing: they measure different things

NPS, CSAT, and CES are not competitors. Each is one way to measure customer satisfaction, and they answer different questions at different points in the customer journey.

• NPS is relational. It asks how a customer feels about your brand overall, and it is tracked at intervals rather than after every interaction.
• CSAT is transactional. It captures satisfaction right at the moment of contact, so it is best sent immediately after a purchase, a support chat, or an onboarding step.
• CES is about effort. Its founding research, published in Harvard Business Review in 2010, found that reducing customer effort predicts loyalty better than trying to delight people.

None of them is sufficient on its own. Used together, chosen by touchpoint, they give a fuller picture of how customers feel, where friction sits, and what to prioritize. The mistake is treating the choice between them as the strategic decision. It is not. The strategic decision is what you do with the reasons behind the number.

NPS vs CSAT vs CES at a glance

When each metric is the right choice

Use NPS when you need a relationship signal you can trend and benchmark. It is the right metric for a quarterly executive view, for comparing regions or product lines, and for tying customer sentiment to growth over time. Be honest about its limits. A 2007 study in the Journal of Marketing found NPS performed no better than other loyalty and satisfaction measures at predicting company growth, and Bain, which created NPS, later introduced an accounting-based "earned growth" measure precisely because the survey number alone was not enough.

Use CSAT when you want fast, specific feedback on a single moment. It is the cleanest read on whether one interaction landed. Satisfaction scores are intuitive to non-analysts and easy to collect at volume, which makes CSAT a good day-to-day operational metric for support and product teams. Its weakness is precision: a 4 out of 5 rarely tells you what would have made it a 5.

Use CES when friction is the thing you are trying to remove. CES is the most predictive of the three for repeat behavior in support contexts. The original research found that 94% of customers reporting low effort said they would repurchase, while 96% of those with high-effort experiences became disloyal. If your risk is customers giving up because something is too hard, CES is the sharpest instrument.

The catch: your score is only as useful as what you know about its drivers

Here is what the standard "use all three" advice leaves out. Whichever metric you track, the number is an outcome. The theme in the open-ended comment next to it is the driver. Scores quantify the result. Themes explain what caused it. If you only watch the score, you are reading the symptom and guessing at the cause.

This is where analyzing the feedback itself, at scale, changes the return on any metric you pick. Thematic reads every open-text response behind the score, groups it into themes, and quantifies each theme's impact on the score, so a CX team can see that stock availability, not price, is what is costing them points.

The proof is in what happens when teams act on the driver instead of the score:

• A NZ home improvement retailer, Mitre10, analyzing 20,000 verbatim comments a month across 84 stores, found that stock availability alone was taking half an NPS point off its overall score. That single driver-level finding shaped a new trade strategy.
• A top-10 US homebuilder lifted closing CSAT from 76.5 to 83.9, a sustained gain of roughly 13 points over three years, by fixing the specific drivers its feedback surfaced rather than chasing the score.
• A national wholesale broadband provider raised episodic NPS by 35 points on one journey and relationship NPS by 6 points at national scale by targeting the themes behind each.
• Atom Bank, the UK digital bank, unified feedback from seven channels across three product lines and tied it to the themes driving its customer scores. Acting on those drivers cut calls on unaccepted mortgage requests by 69%, on savings maturities by 43%, and on device issues by 40%.

In each case the metric was almost incidental. What moved the business was knowing which theme to fix.

How to choose, and what to do next

Work through these questions before you commit to a metric:

1. What decision will this number inform, and how often? Relationship and trend decisions point to NPS. Moment-level operational decisions point to CSAT or CES.
2. Where in the journey are you measuring? Relationship-level points to NPS. A specific interaction points to CSAT or CES.
3. Is your risk dissatisfaction or effort? Dissatisfaction points to CSAT. Friction and giving up point to CES.
4. Can you already see why the number moves? If not, that gap matters more than which metric you picked.

Once you have chosen, put your real effort into the last question. A metric you cannot explain is a metric you cannot improve. Thematic can also generate predicted NPS, CSAT, and effort scores from unstructured feedback such as calls and chats, so touchpoints that never carried a survey still produce a score you can act on.

The short answer

Use NPS for the overall relationship, CSAT for satisfaction with a specific interaction, and CES for how much effort a task took. Choose by touchpoint, not by preference, and do not expect any single number to run your program on its own. The metric you track matters far less than your ability to explain what is driving it. Pick the one that fits the moment, then measure the themes behind it. That is the number that actually moves.